@@ -6,20 +6,8 @@ torchTextClassifiers is a **modular, component-based framework** for text classi
66
77At its core, torchTextClassifiers processes data through a simple pipeline:
88
9- ``` {mermaid}
10- flowchart LR
11- TextInput["Text Input"] --> Tokenizer
12- Tokenizer --> TextEmbedder["Text Embedder"]
13-
14- CatInput["Categorical Features<br/>(Optional)"] --> CatEmbedder["Categorical Embedder"]
15-
16- TextEmbedder --> ClassHead["Classification Head"]
17- CatEmbedder --> ClassHead
18-
19- ClassHead --> Predictions
20-
21- style CatInput stroke-dasharray: 5 5
22- style CatEmbedder stroke-dasharray: 5 5
9+ ``` {thumbnail} diagrams/ttc_architecture.png
10+ :alt: Package Architecture
2311```
2412
2513** Data Flow:**
@@ -195,7 +183,9 @@ The `forward_type` controls how categorical embeddings are combined:
195183
196184Average all categorical embeddings, then concatenate with text:
197185
198- ![ Average and Concatenate] ( diagrams/avg_concat.png )
186+ ``` {thumbnail} diagrams/avg_concat.png
187+ :alt: Average and Concatenate
188+ ```
199189
200190``` python
201191forward_type= CategoricalForwardType.AVERAGE_AND_CONCAT
@@ -209,7 +199,9 @@ forward_type=CategoricalForwardType.AVERAGE_AND_CONCAT
209199
210200Concatenate each categorical embedding separately:
211201
212- ![ Full Concatenation] ( diagrams/full_concat.png )
202+ ``` {thumbnail} diagrams/full_concat.png
203+ :alt: Full Concatenation
204+ ```
213205
214206``` python
215207forward_type= CategoricalForwardType.CONCATENATE_ALL
@@ -283,8 +275,9 @@ head = ClassificationHead(linear=custom_head)
283275
284276## Complete Architecture
285277
286- ![ Complete Architecture] ( diagrams/NN.drawio.png )
287- * Complete model architecture showing all components*
278+ ``` {thumbnail} diagrams/NN.drawio.png
279+ :alt:
280+ ```
288281
289282### Full Model Assembly
290283
0 commit comments